Releases: instructlab/instructlab
Releases · instructlab/instructlab
v0.16.1
v0.16.0
Breaking Changes
- Validate the existence of the model specified in --model by @cdoern in #1102
-
This change requires either rerunning
ilab init
or manually editing yourconfig.yaml
and replacing the followingmodel: merlinite-7b-lab-Q4_K_M
should become
model: models/merlinite-7b-lab-Q4_K_M.gguf
-
Other Changes
- Change mergify automation to require 2 approving reviews for auto-merge by @nathan-weinberg in #1147
- Fixes for linter in lab.py by @alimaredia in #1162
- Fix PyPI release issues by @tiran in #1163
- ci: fix typo in actionlint workflow by @nathan-weinberg in #1168
- mergify: Automate a label for CI failures by @russellb in #1169
- Use quantize binary from instructlab_quantize by @tiran in #1017
- feat: friendlier 'make' by @leseb in #1156
- validation: Confirm yaml file contents are an object by @bjhargrave in #1171
- docs: make document structure more consistent by @nathan-weinberg in #1175
- Docs: Making fixes in README by @kelbrown20 in #1173
- ci: add 'one-approval' label on PR that has one approval by @leseb in #1178
- README.md: Add a build note for Intel Macs by @russellb in #1191
- Create issue template for nominating members for teams by @nathan-weinberg in #1189
- docs: update habana-gaudi.md by @eltociear in #1177
- Add
ilab sysinfo
command by @tiran in #1154 - fix: add chat module to linting and fix errors by @nathan-weinberg in #1188
- fix: update make help text by @nathan-weinberg in #1172
- More efficiently use diskspace by @derekhiggins in #1012
- e2e: Make use of
--non-interactive
forilab init
by @russellb in #1185 - fix: do not fail when sending 'exit' to the chat by @leseb in #1195
- Update llama_cpp_python version to 0.2.75 by @alimaredia in #1161
- Fix CUDA path to account for update by @russellb in #1201
- Docs: Fix converted output by @hickeyma in #1210
- ci: add a ci validation for notebooks by @leseb in #1202
- Add ARM64 support to the Containerfile example by @nerdalert in #987
- use actual converted file name by @makelinux in #1193
- feat: redirect llama_cpp server stderr/stdout to a logger by @leseb in #1179
- Extract model_family from model_path automatically by default. by @makelinux in #1174
- docs: Update install for pypi package by @hickeyma in #1214
- docs: update release documentation by @nathan-weinberg in #1207
- Enforce maximum length in generate seed examples by @bbrowning in #1069
- fix ilab chat run /h command failed issue by @chudegao in #1237
New Contributors
- @eltociear made their first contribution in #1177
- @chudegao made their first contribution in #1237
Full Changelog: v0.15.1...v0.16.0
v0.15.1
v0.15.0
What's Changed
- Docs: Add install python packages on mac by @kelbrown20 in #982
- Fix markdown-lint issues that snuck in by @russellb in #1037
- Use mergify for automating GitHub PR labels by @russellb in #1036
- Ensure we found some knowledge by @derekhiggins in #928
- Enable PyPI publishing and sigstore signing by @tiran in #922
- Disable push to Test PyPI on merge by @tiran in #1045
- --hf-token flag is not used in .14 release by @cdoern in #1047
- Run CI jobs on release branches by @russellb in #1049
- mergify: Allow merging on release branches by @russellb in #1052
- Automatically apply a backport label to release branch PRs by @russellb in #1056
- Chunk docs on spaces for really long lines by @bbrowning in #969
- Add help string to hf-token flag in download by @hickeyma in #1061
- fix: Add hf-token default to fix checking error by @hickeyma in #1064
- fix: remove useless nc command by @leseb in #978
- Ban import of yamllint by @tiran in #1073
- README.md: s/Instruct-Lab/InstructLab/ by @russellb in #1078
- feat: use labels to trigger backports by @leseb in #1074
- Fix container build on machines without avx by @n1hility in #1077
- mergify: Rename backport label to release-branch by @russellb in #1081
- docs: update
STABLE.md
location and content by @nathan-weinberg in #1057 - linux-train: Skip convert to gguf when --4-bit-quant is used by @russellb in #1058
- Fix broken md linting in schema submodule by @hickeyma in #1075
- Update the --model arg by @spzala in #1083
- fix link to "Testing the fine-tuned model" by @makelinux in #1087
- feat: add a welcome message to our server root by @leseb in #1095
- feat: run functional test on MacOS by @leseb in #1039
- ci: fix macos job by @leseb in #1100
- Don't fail if multiprocessing is already set to spawn; fix issues from lack of cleanup by @tiran in #1050
- docs: replace 0.13.0 demo badge with current release badge by @nathan-weinberg in #1099
- Minor update to README.md by @jimmysjolund in #1104
- docs: missing space by @booxter in #1101
- Introduce an end-to-end CI job on Linux by @stefwalter in #1016
- Add Bug and Feature templates for new issues by @nathan-weinberg in #1105
- Document release strategy for Instructlab CLI by @alimaredia in #985
- Do test builds of the cuda image in CI by @russellb in #1086
- chore: clean up intermediate output during MacOS convert by @jaideepr97 in #1097
- Use fine-grained permissions for PyPI by @tiran in #1113
- fix four_bit_quant VRAM check for cuda by @maxamillion in #953
- Docs: Improve instructions for hardware acceleration by @tiran in #1115
- fix: cleaner message when generate dir is empty by @leseb in #1088
- minor change by @lhawthorn in #1119
- notebook: open PR link by @makelinux in #1120
- fix: return model name from the server when chatting by @leseb in #935
- Intel Gaudi / Habana Labs HPU support by @tiran in #966
- add "Open the notebook in Colab" by @makelinux in #1116
- Revert "Intel Gaudi / Habana Labs HPU support" by @russellb in #1122
- beautify link to LoRA by @makelinux in #1117
- tox: add -e docs target to generate man pages by @booxter in #1032
- Add option to include max_tokens for the ilab chat by @luis5tb in #1020
- generate: warn when llama_cpp has not hardware acceleration by @tiran in #1059
- Do not cache llama-cpp-python builds by @tiran in #1125
- Reapply "Intel Gaudi / Habana Labs HPU support" by @tiran in #1128
- fix: do not spellcheck on venv and taxonomy by @leseb in #1130
- Clean exit from the chat.py by @jjasghar in #1131
- Use pytest instead of unittest.TestCase by @tiran in #1135
- chat: Ensure temp server is cleaned up after quit command by @russellb in #1136
- Refactor CliRunner tests by @tiran in #1137
- Fix a minor typo in the troubleshooting guide by @orpiske in #1140
- e2e: Add timestamps to the test script by @russellb in #1139
- docs: add inline mermaid diagram of basic workflow by @makelinux in #1089
- ilab: Document how to enable tab completion by @russellb in #1142
- Issue 1132 - Suggested improvements to "Welcome to the InstructLab CLI" by @adahms in #1133
- fix: handle incorrect log level by @leseb in #1092
- README.md: Fix a typo by @russellb in #1149
- fix: add unit test for config log level by @leseb in #1155
- diff: Rework --quiet logic and handle file without needing --quiet by @bjhargrave in #959
- chat: fix AttributeError crash on -qq with short context by @booxter in #949
- Fix use of
typing.Optional
in config parser by @tiran in #1158 - Add option for install packages on mac by @kelbrown20 in #1110
New Contributors
- @makelinux made their first contribution in #1087
- @jimmysjolund made their first contribution in #1104
- @maxamillion made their first contribution in #953
- @luis5tb made their first contribution in #1020
- @orpiske made their first contribution in #1140
- @adahms made their first contribution in #1133
Full Changelog: v0.14.1...v0.15.0
v0.14.1
v0.14.0
Upgrade Notes
If you are running ilab
straight from a git clone, you must initialize git submodules for this release to work as intended.
git submodule update --init
What's Changed
- Adds contributor ladder file to CLI repo by @stevsmit in #842
- cli-troubleshooting:updates ts and adds resource links by @JoeAldinger in #838
- Restrict server to one client by @derekhiggins in #804
- doc: Remove TODOs from troubleshooting doc by @hickeyma in #871
- (tests) Add tests for knowledge docs by @anik120 in #849
- chore: Remove deprecated files by @hickeyma in #863
- docs: use intro to lab from taxonomy + link&citation by @xukai92 in #875
- chore: Update the license content for included code by @hickeyma in #873
- fix: replace yamllint library install with exec call by @nathan-weinberg in #876
- Update notice comments to single line SPDX-License-Identifier by @bjhargrave in #852
- switch lab coat emoji to dog emoji by @cdoern in #886
- Fix TypeError during
ilab generate
by @tiran in #893 - Docs minor fixes by @booxter in #894
- Reflect new repo name by @russellb in #887
- docs: add demo slides by @xukai92 in #881
- Rename cli package to instructlab by @tiran in #882
- Improve ROCm container and documentation by @tiran in #658
- ci: fix labeler permissions and identation by @nathan-weinberg in #917
- changing default model to instructlab model by @mairin in #921
- provide vuln reporting mechanism by @lhawthorn in #934
- remove remaining mentions of ibm hf repo by @alimaredia in #943
- Unbreak CI and add tox -e spellcheck by @tiran in #955
- Reduce number of discarded generated data samples by @bbrowning in #919
- Update documented training durations by @derekhiggins in #930
- Remove the interaction test about "Barter Town" by @derekhiggins in #927
- Use 'spawn' multiprocessing method instead of fork by @tiran in #957
- tests: fix test file system not isolated by @booxter in #962
- Update default taxonomy repo by @derekhiggins in #915
- validation: Improve error message for too few examples provided by @bjhargrave in #961
- Use matrix configuration for tox by @tiran in #954
- Use pytest fixture to mock mlx package by @tiran in #952
- tests: fix input data file contents by @booxter in #945
- spell: don't be greedy when ignoring backtick fences by @booxter in #944
- fix: ignore SIGINT when running the temp server by @leseb in #925
- CODEOWNERS: Drop file by @russellb in #938
- Use default context for ensure_server by @tiran in #970
- tests: Remove redundant __init__s by @booxter in #946
- Doc: Document key pair size in cli by @kelbrown20 in #971
- Replace Black with Ruff by @tiran in #923
- Restore the note about regen.json by @derekhiggins in #929
- ci: Update labeler.yml with additional files/labels by @nathan-weinberg in #986
- fix: openai and http log level by @leseb in #924
- fix: 'ilab init --non-interactive' does not ask questions anymore by @leseb in #926
- fix: avoid printing errors from the server in the chat by @leseb in #939
- fix: force aspell spellchecker by @leseb in #995
- Prefer https over git and ssh protocols for repo cloning. by @igsilya in #988
- docs: add additional badges to README by @nathan-weinberg in #991
- chore: update chat greeting to InstructLab by @nathan-weinberg in #992
- ci: add stale issue/PR bot by @nathan-weinberg in #1000
- Test with multiple Python versions by @tiran in #984
- Add mergify for PR automation by @russellb in #940
- containers: cuda contianer needs libcudann8 and nvidia-driver-NVML by @stefwalter in #1018
- mergify: Fix config for changes to github workflows by @russellb in #1021
- ci: add markdown linter by @nathan-weinberg in #1013
- containers: Add nvidia-driver-cuda-libs to cuda image by @russellb in #1023
- Fix llama_cpp cuda usage in Containerfile by @n1hility in #1025
- Use schema repository via submodule by @bjhargrave in #913
- feat: ping PR author on merge conflicts by @leseb in #1027
- mergify: Update to latest set of test jobs by @russellb in #1030
- Support serve + chat + generate with Mixtral teacher model by @cdoern in #1002
- Define a relaxed schema as v1 and move current schema to v2 by @bjhargrave in #1009
- Fix tox -e spellcheck by @tiran in #1014
- Make
ilab
able to download safetensors by @cdoern in #1033 - github: Add a test-workflow-complete job by @russellb in #1031
- Follow up to safetensor support by @cdoern in #1035
New Contributors
- @lhawthorn made their first contribution in #934
- @igsilya made their first contribution in #988
- @n1hility made their first contribution in #1025
Full Changelog: v0.13.1...v0.14.0
v0.13.1
v0.13.0
What's Changed
- fix: clarify initialize lab instructions by @mairin in #825
- Adds maintainers.md file and links to community repo maintainers file by @stevsmit in #830
- Adds security advisory link to community repo by @stevsmit in #832
- Get system prompt from a single source by @derekhiggins in #824
- Add TLS settings for connecting to hosted model by @hickeyma in #828
- Use the taxonomy schema to validate taxonomy yaml files by @bjhargrave in #776
- Adds governance file and links to community repo's governance document by @stevsmit in #839
- cli: Fix regression in call to list_models() by @russellb in #840
- cli-readme:style guideline updates by @JoeAldinger in #829
- fix: regression in train command when single source system prompt added by @hickeyma in #843
- schema: Remove code no longer needed since we schema validate yaml by @bjhargrave in #841
New Contributors
- @stevsmit made their first contribution in #830
- @JoeAldinger made their first contribution in #829
Full Changelog: v0.12.1...v0.13.0
v0.12.1
What's Changed
- Fix broken URL in Readme by @vishnoianil in #794
- Don't duplicate training data by @derekhiggins in #763
- chat: Add --endpoint-url and --api-key options by @russellb in #795
- Use setuptools_scm to auto-generate version by @tiran in #737
- adding attribution to original codebase by @abhi1092 in #799
- diff: Fix lint problem output by @bjhargrave in #805
- Reduce some log level for k-doc fetching logic by @anik120 in #806
New Contributors
- @vishnoianil made their first contribution in #794
Full Changelog: v0.12.0...v0.12.1
v0.12.0 Knowledge Support
What's Changed
- Feature/knowledge doc git by @shahfasal in #768
- Multi-doc support for knowledge contributions by @aartij22 in #771
- fix: run the temp server with an available port by @leseb in #744
- fix: Improve diff and taxonomy YAML logging by @rafvasq in #756
- Allow lab test to run without a training run by @russellb in #765
- fix: properly propagate errors when temp server fails to start by @leseb in #758
- fix: reduce client timeout while attempting to connect to server by @leseb in #742
- Introduce Makefile that automates container builds by @alinaryan in #622
- Add test state badge to README.md by @khaledsulayman in #779
- Allow multiple file patterns for knowledge docs by @anik120 in #781
- add langchain for chunking by @aartij22 in #772
- pin llama_cpp_python to 0.2.55 by @anik120 in #789
- Knowledge doc submissions will have schema with
patterns
by @anik120 in #791 - Add cov-report term to see coverage when tests run by @markstur in #793
- Knowledge doc chunking follow up by @anik120 in #792
New Contributors
- @shahfasal made their first contribution in #768
- @aartij22 made their first contribution in #771
- @alinaryan made their first contribution in #622
Full Changelog: v0.11.0...v0.12.0